> ## Documentation Index
> Fetch the complete documentation index at: https://docs.trilo.chat/llms.txt
> Use this file to discover all available pages before exploring further.

# Custom AI Coworkers

> Create AI assistants tailored to your team's specific needs

While Trilo provides built-in AI coworkers, you can create custom AI coworkers that are specialized for your team's unique needs. Custom coworkers can have specific personalities, expertise areas, and behaviors.

## Why Create Custom Coworkers?

Custom AI coworkers let you:

* Match your brand voice and tone
* Specialize in your domain knowledge
* Create consistent responses for your use cases
* Define specific behaviors and constraints
* Build AI that understands your context

## Creating a Custom Coworker

### Getting Started

<Steps>
  <Step title="Go to AI Coworkers">
    Navigate to the **AI Coworkers** section in settings.
  </Step>

  <Step title="Click Create">
    Click **Create AI Coworker** or **+ New**.
  </Step>

  <Step title="Configure basics">
    Set up name, avatar, and description.
  </Step>

  <Step title="Define behavior">
    Write instructions for how the AI should behave.
  </Step>

  <Step title="Save">
    Save your custom coworker. It's now available to your team.
  </Step>
</Steps>

### Basic Settings

#### Name

Choose a clear, memorable name:

* Reflects the coworker's purpose
* Easy to identify in lists
* Professional and appropriate

**Examples:**

* "Brand Voice Editor"
* "Customer Support Assistant"
* "Technical Documentation Helper"

#### Avatar

Set a visual identity:

* Upload a custom image
* Use generated avatars
* Choose colors that fit your brand

#### Description

Write a brief description that helps users understand:

* What this coworker does
* When to use it
* Its areas of expertise

**Example:**

```
Helps write customer-facing communications in our
brand voice. Specializes in support emails, product
announcements, and social media content.
```

## Defining Behavior

### Base Prompt

The base prompt tells the AI how to behave. This is the most important configuration.

**Structure your prompt to include:**

1. **Role definition**: Who is this AI?
2. **Expertise**: What does it know?
3. **Tone and style**: How should it communicate?
4. **Guidelines**: Rules it should follow
5. **Constraints**: What it should avoid

### Example Base Prompt

```
You are a Customer Support Assistant for Acme Corp.

**Role:** Help team members draft responses to customer
inquiries, ensuring consistent and helpful communication.

**Expertise:**
- Product knowledge for Acme's software suite
- Common customer issues and solutions
- Support best practices

**Tone:**
- Friendly and professional
- Empathetic to customer concerns
- Clear and concise
- Solution-oriented

**Guidelines:**
- Always acknowledge the customer's concern first
- Provide clear next steps
- Offer additional resources when relevant
- End with a helpful closing

**Avoid:**
- Technical jargon customers won't understand
- Blaming the customer
- Making promises we can't keep
- Sharing internal processes
```

### Writing Effective Prompts

<AccordionGroup>
  <Accordion title="Be specific">
    Vague instructions lead to inconsistent results. Clearly state what you want.
  </Accordion>

  <Accordion title="Include examples">
    Show the AI what good output looks like.

    ```
    Example response style:
    "Hi [Name], I completely understand how frustrating this must be.
    Let me help you resolve this right away..."
    ```
  </Accordion>

  <Accordion title="Define boundaries">
    Tell the AI what NOT to do as clearly as what to do.
  </Accordion>

  <Accordion title="Test and iterate">
    After creating, test with real scenarios. Adjust the prompt based on results.
  </Accordion>
</AccordionGroup>

## Advanced Configuration

### Skills and Capabilities

Enable or configure specific capabilities:

| Capability            | Description                                |
| --------------------- | ------------------------------------------ |
| **Task Creation**     | Can create tasks from conversations        |
| **Image Generation**  | Can generate images (if available)         |
| **Document Analysis** | Can process uploaded files                 |
| **Web Search**        | Can search for information (if configured) |

### Temperature Setting

Control how creative vs. consistent the AI is:

* **Lower (0.1-0.3)**: More consistent, focused responses
* **Medium (0.5-0.7)**: Balanced creativity and consistency
* **Higher (0.8-1.0)**: More creative, varied responses

**When to adjust:**

* Support responses: Lower temperature for consistency
* Creative writing: Higher temperature for variety
* Technical answers: Lower temperature for accuracy

### Token Limits

Set maximum response length:

* Shorter limits for concise answers
* Longer limits for detailed content
* Balance between thoroughness and brevity

## Use Case Examples

### Customer Support Assistant

```
**Name:** Support Helper

**Description:** Drafts customer support responses
following our guidelines and tone.

**Base Prompt:**
You help draft customer support emails. Follow these principles:
- Always empathize first
- Provide clear solutions
- Use simple, jargon-free language
- Include relevant help center links
- Sign off warmly

Avoid: defensive language, technical terms,
acknowledging bugs not yet announced publicly.
```

### Content Editor

```
**Name:** Brand Voice Editor

**Description:** Reviews and edits content to match
our brand voice and style guide.

**Base Prompt:**
You are an editor who ensures all content matches
our brand voice.

Our voice is:
- Confident but not arrogant
- Friendly and approachable
- Clear and concise
- Optimistic and helpful

When reviewing content:
1. Check for brand voice alignment
2. Suggest specific improvements
3. Flag any issues
4. Provide the edited version
```

### Onboarding Guide

```
**Name:** Onboarding Buddy

**Description:** Helps new team members learn
our processes and tools.

**Base Prompt:**
You help new team members get up to speed.
You're friendly, patient, and thorough.

Knowledge areas:
- Company tools and how to use them
- Team processes and workflows
- Where to find documentation
- Who to ask for what

Always:
- Welcome questions, no matter how basic
- Point to relevant resources
- Suggest logical next steps
- Encourage asking follow-up questions
```

## Managing Custom Coworkers

### Editing

To update a custom coworker:

1. Go to AI Coworkers in settings
2. Find your custom coworker
3. Click **Edit**
4. Make changes
5. Save

Changes apply to new conversations.

### Duplicating

Create variations based on existing coworkers:

1. Find the coworker to copy
2. Click **Duplicate**
3. Modify as needed
4. Save with a new name

### Deleting

Remove coworkers no longer needed:

1. Find the coworker
2. Click **Delete**
3. Confirm deletion

<Warning>
  Deleting a custom coworker removes it permanently. Existing conversation history remains but new conversations can't be started.
</Warning>

## Sharing Custom Coworkers

### Workspace Availability

Custom coworkers can be available to:

* **Entire workspace**: Everyone can use it
* **Specific projects**: Only in certain projects
* **Specific roles**: Only certain team members

### Permissions

Control who can:

* **Use**: Start conversations with the coworker
* **Edit**: Modify the coworker's configuration
* **Delete**: Remove the coworker

## Best Practices

<AccordionGroup>
  <Accordion title="Start simple">
    Begin with a basic configuration. Add complexity as you learn what works.
  </Accordion>

  <Accordion title="Test thoroughly">
    Before rolling out, test with various scenarios to ensure consistent behavior.
  </Accordion>

  <Accordion title="Gather feedback">
    Ask team members how well the coworker performs. Use feedback to improve.
  </Accordion>

  <Accordion title="Document purpose">
    Write clear descriptions so team members know when to use each coworker.
  </Accordion>

  <Accordion title="Maintain over time">
    Review and update coworkers as your needs evolve. Outdated prompts give outdated results.
  </Accordion>
</AccordionGroup>

## Troubleshooting

<AccordionGroup>
  <Accordion title="Coworker not following instructions">
    * Make instructions more explicit
    * Add examples of correct behavior
    * Remove conflicting guidelines
    * Test with the prompt to refine
  </Accordion>

  <Accordion title="Responses are inconsistent">
    * Lower the temperature setting
    * Be more specific in the prompt
    * Add constraints and boundaries
    * Include more examples
  </Accordion>

  <Accordion title="Coworker is too verbose">
    * Add instruction: "Be concise"
    * Set lower token limits
    * Specify desired response length
    * Include examples of ideal length
  </Accordion>
</AccordionGroup>

## Next Steps

<CardGroup cols={2}>
  <Card title="AI Capabilities" icon="wand-magic-sparkles" href="/ai-coworkers/ai-capabilities">
    Explore what AI coworkers can do.
  </Card>

  <Card title="Chatting with AI" icon="comment" href="/ai-coworkers/chatting-with-ai">
    Learn effective AI communication.
  </Card>
</CardGroup>